Relative to healthy adults, depressed individuals typically show excellent memory for negative material but poor memory for positive material. Furthermore, depression impairs recollection?the ability to retrieve vivid, contextual details about an event. These abnormalities trouble patients and appear to prolong depressive episodes, but they are not well understood. Therefore, this proposal will use multi-modal neuroimaging and computational modeling to investigate the encoding and retrieval of emotional memories in depressed adults. To investigate categorical effects of depression, electroencephalogram (EEG)/event-related potential (ERP) data will be collected from 64 unmedicated adults with MDD and 64 healthy controls (n = 128). To investigate dimensional effects of depression, functional magnetic resonance imaging (fMRI) data will be acquired from adults selected for minimal, mild, moderate, or severe depressive symptoms (n = 36). On Day 1, the participants will study negative and positive words in the context of two encoding tasks. On Day 2, they will return for a recognition memory test in which the ?old? encoded words will be presented with similar ?new? words. When a participant recognizes an old word, source memory (recollection) will be tested by asking which task the word was encoded with. Day 2 will include exposure to acute stress, to potentiate emotional biases. This comprehensive design will support three aims. Aim 1 will use EEG/ERP to test the hypothesis that MDD blunts cortical responses to positive vs. negative stimuli at encoding and retrieval. We expect ERPs linked to memory formation and retrieval to be reduced for positive material, but not negative material, in adults with MDD vs. controls. Moreover, we expect such effects to be exaggerated after stress exposure. Importantly, the EEG/ERP methodology cannot detect activity in subcortical brain regions important for memory, such as the amygdala. Therefore, Aim 2 will use fMRI to test the hypothesis that depressive severity correlates with activation in subcortical structures that support retrieval. We expect that as depressive severity increases, activation of the amygdala, hippocampus, and parietal cortex to negative memory probes will increase. By contrast, activation of the striatum, hippocampus, and parietal cortex to positive memory probes should decrease. Finally, to gain insight into the underlying mechanisms that support memory, Aim 3 will use the HDDM to reveal the impact of depression on decision-making at retrieval. The Hierarchical Drift Diffusion Model (HDDM) is a computational model that can estimate the evidence accumulation process that enables us to choose between two options (e.g., old vs. new). We predict that the speed of evidence accumulation?drift rate?will be reduced for positive, but not negative, memory probes in depressed adults. Moreover, increased depression is expected to weaken relationships between drift rate and EEG/fMRI signals that support memory for positive material, but it should strengthen such relationships for negative material. This combination of computational modeling and multi-modal imaging will yield new insight into memory deficits in depression.